Remember, the job of the ADC is to convert an audio signal from the analog to the digital realm. Every 1/44,100th of a second when the ADC records and stores the amplitude levels, it stores them as numbers. These numbers are stored digitally in a computer’s (or recording device’s) hardware memory as bits [1]. It turns out that with n bits you can store 2n values. A single bit (21) can offer just two values – a 1 or 0 – to which to assign an amplitude sample from a signal. This means that it records the voltage all the way on (1) or all the way off (0). Obviously, one-bit amplitude resolution is pretty course! Even four bits (24), offering 16 different levels to which the voltage may be assigned, is far too crude a resolution to describe with any detail an audio signal. The figure below shows a continuous (analog) sine wave and the same wave sampled digitally in four bit resolution.
The dynamic range (loudest to softest sound in a system) is reflected in the vertical aspect of the graph, which is divided into the 16 levels available with four bits. The way actual amplitudes along the original sine wave are forced to the nearest of the 16 levels is called quantizing (rounding). The difference between any of these quantized amplitude values, and the actual amplitude of the sine wave at that time, is digital noise (sometimes called quantizing noise).
Obviously the more bits a device can devote to describing amplitudes of samples, the truer the sound and the less digital noise. [2]
[1] Bit is short for binary digit. Eight bits form a byte.
[2] Without going into the gory details, it turns out that for every bit used to store amplitude sample values in a digital system, you reduce the noise by 6 dBs (decibels). An 8-bit recording device has a dynamic range of 48 dBs; the ratio of the loudest sound it can represent is 48 dBs (8 bits x 6 dBs) above the noise floor. For a 16-bit system, the loudest sound is a full 96 dBs above the noise floor. With 16-bits, noise created by quantizing is basically unnoticeable.
The dynamic range (loudest to softest sound in a system) is reflected in the vertical aspect of the graph, which is divided into the 16 levels available with four bits. The way actual amplitudes along the original sine wave are forced to the nearest of the 16 levels is called quantizing (rounding). The difference between any of these quantized amplitude values, and the actual amplitude of the sine wave at that time, is digital noise (sometimes called quantizing noise).
Obviously the more bits a device can devote to describing amplitudes of samples, the truer the sound and the less digital noise. [2]
SAMPLE RATE & BIT DEPTH EXPLAINED
A More Thorough Explanation of Sample Resolution (Bit Depth)
[1] Bit is short for binary digit. Eight bits form a byte.
[2] Without going into the gory details, it turns out that for every bit used to store amplitude sample values in a digital system, you reduce the noise by 6 dBs (decibels). An 8-bit recording device has a dynamic range of 48 dBs; the ratio of the loudest sound it can represent is 48 dBs (8 bits x 6 dBs) above the noise floor. For a 16-bit system, the loudest sound is a full 96 dBs above the noise floor. With 16-bits, noise created by quantizing is basically unnoticeable.